Grafton Sciences is founded on a simple idea: the history of progress is a story of exploration and learning. Learning refines what is known and guides where to go next. Exploration uncovers what is missing. Every major advance in science and engineering has relied on this interplay.

The next jump in intelligence requires a general substrate for exploration and learning.

Modern AI architectures already provide a strong substrate for learning. What we lack is an equivalent substrate for exploration in the physical world. Without a way to freely query the physical world, the pace of progress will be constrained.

We are building the missing layer.

We are training models and building the factories they need for general purpose exploration in the real world. Our goal is to build a substrate robust enough to enable experimentation, engineering, or manufacturing of anything, a kind of general physical ability.

Our aim is to pair this ‘open-ended substrate’ with advanced learning systems so that exploration and learning reinforce each other at scale.

We believe that building an unconstrained platform for physical exploration is a key step toward much more capable AI systems and a new rate of scientific progress.

Along the way, we expect to produce and publish advances across a range of scientific domains.

Grafton Sciences Team
4.15.2025